Preferential-looking attention prototype
Two videos are shown side by side: geometric patterns and social interaction, with left/right placement randomised. The webcam estimates where the child is looking, and the app reports the share of time spent toward each video.
Without uploaded files, a built-in scripted 30-second pair is used: moving geometric patterns and two cartoon characters playing together, six 5-second scenes each. The geometric video is generated with a different variation for each session (the same variation for both passes of a session) and is checked to match the social video in brightness, screen coverage and motion before it is shown. The social video is always the same (coded-v1). They are cartoon animations, not real-person video, and are not clinically validated. Videos play muted.